JP2023113171A - Voice processing unit, voice processing method, voice processing program and voice processing system - Google Patents

Abstract

To suppress erroneous detection of voice recognition.SOLUTION: A voice processing unit 10 includes: a voice acquisition part 20; a determination part 22; a voice processing part 24; and a switching part 26. The voice acquisition part 20 acquires a voice signal from a microphone MC for collecting voice in space. The determination part 22 determines whether or not a level of a reference signal which is a reproduction signal reproduced from a speaker SP for emitting sound in the space is equal to or higher than a threshold value. The voice processing part 24 outputs a removal signal obtained by removing a voice component of the reference signal from the voice signal, to a voice recognition part 40 as an output signal. . When the level of the reference signal is determined to be equal to or higher than the threshold value, the switching part 26 outputs a replacement signal which is at least one of comfort noise and a mute signal, to the voice recognition part 40 as the output signal instead of the removal signal.SELECTED DRAWING: Figure 3

Description

The present disclosure relates to an audio processing device, an audio processing method, an audio processing program, and an audio processing system.

Speech processing systems are known that process speech recognition commands based on speech uttered by a speaker. For example, the first speech recognition unit recognizes the sound picked up by the microphone, and the second speech recognition unit recognizes the sound output from the speaker. A configuration is disclosed that stops recognition by the first speech recognition unit when the speech recognized by the second speech recognition unit includes a speech recognition command (for example, see Patent Document 1).

Japanese Patent No. 6225920

However, in the prior art, when the voice picked up by the microphone contains noise components such as residual echo components that cannot be completely removed by the echo canceller, erroneous detection of voice recognition may occur. That is, in the prior art, it was sometimes difficult to suppress erroneous detection of voice recognition.

An object of the present disclosure is to provide a speech processing device, a speech processing method, a speech processing program, and a speech processing system capable of suppressing erroneous detection of speech recognition.

A speech processing device according to an aspect of the present disclosure includes a speech acquisition unit, a determination unit, a speech processing unit, and a switching unit. The audio acquisition unit acquires an audio signal from a microphone that picks up spatial audio. The determination unit determines whether or not the level of a reference signal, which is a reproduction signal reproduced from a speaker emitting sound in the space, is equal to or higher than a threshold. The speech processing unit outputs a removal signal obtained by removing the speech component of the reference signal from the speech signal to the speech recognition unit as an output signal. The switching unit outputs a replacement signal, which is at least one of comfort noise and a mute signal, as the output signal to the speech recognition unit in place of the removal signal when the level of the reference signal is determined to be equal to or higher than the threshold. .

According to the present disclosure, erroneous detection of speech recognition can be suppressed.

FIG. 1 is a diagram showing an example of a schematic configuration of a speech processing system 1 of this embodiment. FIG. 2 is a hardware configuration diagram of an example of the audio processing device. FIG. 3 is a block diagram showing an example of the configuration of the audio processing device. FIG. 4 is a flowchart showing an example of the flow of information processing executed by the speech processing device 10 of this embodiment.

Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings as appropriate. However, more detailed description than necessary may be omitted. It should be noted that the accompanying drawings and the following description are provided for a thorough understanding of the present disclosure by those skilled in the art and are not intended to limit the claimed subject matter.

FIG. 1 is a diagram showing an example of a schematic configuration of a speech processing system 1 of this embodiment.

The speech processing system 1 is a system for recognizing speech in space. In this embodiment, a case where the space is the space in the vehicle interior of the vehicle 2 will be described as an example. In addition, in the present embodiment, a form in which the voice processing system 1 is mounted in the vehicle 2 will be described as an example. Note that the space is not limited to the interior of the vehicle 2 .

The voice processing system 1 includes a microphone MC, a speaker SP, a voice processing device 10 , a sound source device 30 , a voice recognition section 40 , an electronic device 50 and a display 60 . Microphone MC, speaker SP, speech recognition unit 40, display 60, and speech processing device 10 are communicably connected. The voice processing system 1 may be configured to include at least the microphone MC, the speaker SP, the voice processing device 10, and the voice recognition section 40. FIG.

A microphone MC picks up sounds in the space. In this embodiment, the microphone MC picks up at least the sound of the space inside the vehicle 2 . In this embodiment, an example will be described in which the microphone MC is provided in the vicinity of the driver's seat, which is the seat of the driver hm1 of the vehicle 2 . Therefore, in the present embodiment, the microphone MC picks up the voice including at least the voice component uttered by the driver hm1.

The vehicle 2 may be configured to have a plurality of microphones MC. In this case, it is preferable that these multiple microphones MC are arranged at mutually different positions in the cabin of the vehicle 2 . Specifically, for example, the microphones MC may be arranged near the seats of the driver hm1, the passenger hm2, the passenger hm3, and the passenger hm3 of the vehicle 2, respectively. In the present embodiment, a mode in which one microphone MC is provided in the vehicle 2 will be described as an example.

The microphone MC may be either a directional microphone or an omnidirectional microphone. The microphone MC may be either a small MEMS (Micro Electro Mechanical Systems) microphone or an ECM (Electret Condenser Microphone). Microphone MC may be a beamforming microphone. For example, the microphone MC may be a microphone array that has directivity in a specific direction and is capable of picking up sound in the directional direction.

The microphone MC outputs an audio signal of the collected audio to the audio processing device 10 . The voice processing device 10 is provided in association with the microphone MC. Therefore, when the audio processing system 1 is configured to include multiple microphones MC, the audio processing system 1 may be configured to include multiple audio processing devices 10 corresponding to each of the multiple microphones MC. In this embodiment, an example will be described in which the audio processing system 1 includes one microphone MC and one audio processing device 10 communicably connected to the microphone MC.

The speaker SP emits sound to the same space as the sound pickup target space of the microphone MC. In this embodiment, the speaker SP emits sound at least to the space inside the vehicle 2 .

In the present embodiment, as an example, a form in which four speakers SP of speakers SP1 to SP4 are arranged in the vehicle interior of the vehicle 2 will be described. Note that the audio processing system 1 may be configured to include at least one speaker SP, and the number and arrangement positions of the speakers SP are not limited. In this embodiment, a speaker SP1, a speaker SP2, a speaker SP3, and a speaker SP4 are arranged near the seats of the driver hm1, the passenger hm2, the passenger hm3, and the passenger hm3 in the cabin of the vehicle 2, respectively. The form will be described as an example. Note that when these speakers SP1 to SP4 are collectively described, they are simply referred to as the speaker SP.

The speaker SP is electrically connected to the sound source device 30 . The speaker SP emits sound represented by the reproduction signal received from the sound source device 30 . A reproduction signal is a signal output from the sound source device 30 to the speaker SP. The speaker SP emits sound corresponding to the reproduction signal received from the sound source device 30 . Specifically, the speaker SP emits sound with a volume corresponding to the level of the reproduction signal received from the sound source device 30 . That is, in this embodiment, the level means the level of a signal, and specifically means the volume of sound represented by the signal.

The sound source device 30 is, for example, a radio receiver, a television broadcast device, an audio device, or the like. The radio receiver receives a radio broadcast signal, generates a reproduced signal from the received radio broadcast signal, and outputs the reproduced signal to the speaker SP. In this case, the reproduced signal is, for example, a radio audio signal of radio audio. The television broadcasting apparatus receives a television broadcasting signal, generates a reproduction signal from the received television broadcasting signal, and outputs it to the speaker SP. In this case, the reproduced signal is, for example, a television audio signal of television audio. The audio device outputs a reproduced signal such as an audio signal recorded in a memory or the like to the speaker SP. In this case, the reproduced signal is, for example, an audio signal.

In this embodiment, the sound source device 30 generates 4-channel reproduction signals to use the four speakers SP (speakers SP1 to SP4), and outputs them as reference signals to each of the four speakers SP. Specifically, the sound source device 30 outputs a reference signal 1, which is a reproduced signal, to the speaker SP1, outputs a reference signal 2, which is a reproduced signal, to the speaker SP2, and outputs a reference signal 3, which is a reproduced signal, to the speaker SP3. , the reference signal 4, which is a reproduced signal, is output to the speaker SP4. These reference signals 1 to 4 are reproduction signals output to each of the plurality of speakers SP. When the reference signals 1 to 4 are collectively described, they will simply be referred to as reference signals.

The speech processing device 10 outputs to the speech recognition section 40 an output signal based on the reference signal, which is the speech signal received from the microphone MC and the reproduction signal reproduced from the speaker SP. The details of the audio processing device 10 will be described later.

The speech recognition unit 40 recognizes speech represented by the output signal received from the speech processing device 10 and outputs a signal representing the speech recognition result to the electronic device 50 . For example, the voice recognition unit 40 recognizes voice commands represented by output signals and outputs them to the electronic device 50 . A voice command is a signal for causing the electronic device 50 to execute various processes. Voice commands may also be referred to as voice recognition commands, keywords, wake-up words, and the like.

Electronic device 50 executes processing according to a voice command, which is a signal representing a voice recognition result received from voice recognition unit 40 . For example, the electronic device 50 executes processing for opening and closing windows, processing for driving the vehicle 2, processing for changing the temperature of the air conditioner, processing for changing the volume of the audio device, and the like, based on the voice command. The electronic device 50 is, for example, a car navigation device, an air conditioner, a panel meter, a television, a mobile terminal, a driving device that drives each part of the vehicle 2, and the like.

The display 60 is a display device that displays various information. The display 60 is, for example, various displays provided in the vehicle 2, a head-up display, a display of a car navigation system, a multi-information display provided in a meter of the vehicle 2, a center display capable of accepting audio operations, etc. is. In this embodiment, information is displayed on the display 60 by the audio processing device 10, which will be described later. Note that the display 60 may function as an example of the electronic device 50 .

The speech processing device 10 will be described in detail. First, an example of the hardware configuration of the audio processing device 10 will be described.

FIG. 2 is a hardware configuration diagram of an example of the audio processing device 10. As shown in FIG.

The audio processing device 10 includes a CPU (Central Processing Unit) 11A, a ROM (Read Only Memory) 11B, a RAM 11C, and an I/F 11D, etc., which are interconnected via a bus 11E. It's becoming

The CPU 11A is an arithmetic device that controls the speech processing device 10 of this embodiment. The ROM 11B stores programs and the like for realizing various processes by the CPU 11A. The RAM 11C stores data required for various processes by the CPU 11A. I/F 11D is an interface for transmitting and receiving data.

A program for executing information processing executed by the speech processing apparatus 10 of the present embodiment is provided by being incorporated in the ROM 11B or the like in advance. The program executed by the sound processing device 10 of the present embodiment is a file in a format that can be installed in the sound processing device 10 or in a format that can be executed on CD-ROM, flexible disk (FD), CD-R, DVD ( It may be configured to be provided by being recorded on a computer-readable recording medium such as a Digital Versatile Disk).

Next, the configuration of the audio processing device 10 will be described in detail.

FIG. 3 is a block diagram showing an example of the configuration of the speech processing device 10. As shown in FIG. For the sake of explanation, FIG.

The audio processing device 10 includes an audio acquisition unit 20 , a determination unit 22 , an audio processing unit 24 , a switching unit 26 , a generation unit 28 and an output control unit 29 .

Some or all of the voice acquisition unit 20, the determination unit 22, the voice processing unit 24, the switching unit 26, the generation unit 28, and the output control unit 29 cause a processing device such as the CPU 11A to execute a program, that is, It may be implemented by software, may be implemented by hardware such as an IC (Integrated Circuit), or may be implemented by using both software and hardware. At least one of the voice acquisition unit 20, the determination unit 22, the voice processing unit 24, the switching unit 26, the generation unit 28, and the output control unit 29 is communicably connected to the voice processing device 10 via a network or the like. Alternatively, it may be installed in an external information processing device.

The voice acquisition unit 20 acquires a voice signal from the microphone MC. The audio acquisition unit 20 outputs the acquired audio signal to the audio processing unit 24 .

The determination unit 22 determines whether or not the level of the reference signal, which is the reproduction signal reproduced from the speaker SP, is equal to or higher than the threshold. The level of the reference signal represents the loudness of the sound represented by the reproduced signal, which is the reference signal. As described above, the speaker SP emits sound with a volume corresponding to the level of the reproduced signal received from the sound source device 30 . Therefore, the higher the level of the reference signal, which is the reproduced signal, the higher the volume of the sound emitted from the speaker SP.

The threshold value is equal to or lower than, and close to, the level of the reproduced signal when the level of the reproduced signal is gradually increased and distortion begins to occur in the sound output from the speaker SP according to the reproduced signal. A value may be determined in advance. Also, the threshold value may be a value that coincides with the level of the reproduced signal when the level of the reproduced signal is gradually increased and distortion begins to occur in the sound emitted from the speaker SP according to the reproduced signal. good. Distortion of the sound output from the speaker SP is sometimes referred to as sound distortion.

For example, the determination unit 22 determines a threshold that satisfies the above conditions for each of the plurality of speakers SP1 to SP4.

Then, the determination unit 22 determines that at least one of the levels of the reference signals 1 to 4 received from each of the plurality of speakers SP1 to SP4 is equal to or higher than the threshold value corresponding to each of the speakers SP1 to SP4. Determine whether or not there is

Further, the determination unit 22 sets the minimum value, the average value, or the maximum value of the threshold values that satisfy the above conditions for each of the plurality of speakers SP1 to SP4 as a common threshold value for the plurality of speakers SP1 to SP4. good too. Then, the determination unit 22 determines whether at least one of the levels of the reference signals 1 to 4 received from each of the plurality of speakers SP1 to SP4 is equal to or higher than the threshold set as a common threshold. It may be determined whether

In this embodiment, the determination unit 22 determines that at least one of the levels of the reference signals 1 to 4 received from each of the plurality of speakers SP1 to SP4 corresponds to each of the speakers SP1 to SP4. A form of determining whether or not it is equal to or greater than a threshold will be described as an example.

Note that threshold values corresponding to each of the plurality of speakers SP1 to SP4 may be stored in advance in the memory or the like of the determination unit 22. FIG. Further, the threshold value corresponding to each of the plurality of speakers SP1 to SP4 is appropriately changed according to the type and installation position of the speaker SP provided in the audio processing system 1 within the range that satisfies the above conditions by an operation instruction or the like by the user. It may be possible.

The audio processing unit 24 generates a removal signal by removing the audio component of the reference signal from the audio signal received from the audio acquisition unit 20 .

The audio processing unit 24 removes the audio component of the reference signal, which is the reproduced signal, included in the audio signal received from the audio acquisition unit 20 . The audio processing unit 24 may remove the audio component of the reference signal included in the audio signal using at least one of a known echo canceller and crosstalk canceller.

For example, the audio processing unit 24 has an adaptive filter F, an adaptive filter control unit 24A, and a subtraction unit 24B.

The adaptive filter F is a filter having a function of changing the characteristics of the reference signal. In this embodiment, the adaptive filter F includes adaptive filters F1 to F4. The number of adaptive filters F is appropriately set based on the number of input reference signals and the like.

The adaptive filter control section 24A sets the filter coefficients of the adaptive filters F1 to F4 by a known method according to the removal signal output from the subtraction section 24B. The adaptive filters F1 to F4 output passing signals based on the respective received reference signals 1 to 4 and the set filter coefficients to the subtraction section 24B as subtraction signals. For this reason, in the subtraction unit 24B, a signal obtained by adding the reference signals 1 to 4 output from the adaptive filters F1 to F4 and the passing signals based on the set filter coefficients A subtraction signal is output.

The subtraction unit 24B subtracts the subtraction signal from the audio signal received from the audio acquisition unit 20, thereby performing a removal process of removing the audio component of the reference signal from the audio signal. The subtraction unit 24B outputs the removal signal obtained by the removal processing, that is, the removal signal obtained by removing the audio component of the reference signal from the audio signal to the adaptive filter control unit 24A and the switching unit 26. FIG.

When it is determined that the level of the reference signal is equal to or higher than the threshold, the switching unit 26 converts the replacement signal, which is at least one of the comfort noise and the mute signal, into the output signal instead of the removal signal received from the audio processing unit 24, and converts it into the output signal. 40.

Specifically, when the determination unit 22 determines that the level of the reference signal is equal to or higher than the threshold, the switching unit 26 replaces the removal signal received from the speech processing unit 24 with the replacement signal received from the generation unit 28 for speech recognition. Switch to output to unit 40 .

The generation unit 28 generates a replacement signal that is at least one of the comfort noise and the mute signal, and outputs the replacement signal to the switching unit 26 . A mute signal is a signal whose sound level is "0". A mute signal, in other words, is a signal representing a silence, mute, or no signal (MUTE).

When generating the comfort noise as the replacement signal, the generation unit 28 can generate the comfort noise at a level corresponding to the noise level included in the audio signal at the timing just before the determination unit 22 determines that the noise is equal to or greater than the threshold. preferable. For example, the audio acquisition unit 20 outputs audio signals acquired from the microphone MC to the audio processing unit 24 and the generation unit 28 . The generation unit 28 identifies, by a known method, the noise level included in the audio signal received from the audio acquisition unit 20 at the timing immediately before the determination unit 22 determines that the noise level is equal to or greater than the threshold. Then, the generator 28 generates comfort noise of a level corresponding to the specified noise level. For example, the generation unit 28 generates comfort noise representing the same level of noise level as the specified noise level, that is, the same level of volume.

The level of the output signal output to the speech recognition unit 40 is generated by the generation unit 28 generating, as a replacement signal, comfort noise having a level corresponding to the noise level included in the audio signal at the timing immediately before it is determined to be equal to or greater than the threshold. is suppressed from rapidly changing. For example, when the spatial sound environment fluctuates according to a change in the traveling environment of the vehicle 2 or the like, comfort noise having a level corresponding to the fluctuation of the spatial sound environment is output to the speech recognition unit 40 as a substitution signal. Therefore, when the output signal output to the speech recognition unit 40 switches from the replacement signal to the removal signal or from the removal signal to the replacement signal, the level of the output signal is suppressed from fluctuating abruptly. Therefore, it is possible to suppress deterioration of the speech recognition performance of the speech recognition unit 40 due to sudden changes in the level of the output signal.

Further, the generating section 28 may generate a replacement signal including both the comfort noise and the mute signal and output it to the switching section 26 . For example, the generator 28 generates a replacement signal in which comfort noise and mute signals are alternately arranged. In this case, the generator 28 preferably generates an output signal whose level is adjusted so that the level changes gradually when the comfort noise and the mute signal are switched.

Note that the generation unit 28 may always generate the replacement signal, but it is preferable to generate the replacement signal and output it to the switching unit 26 when the determination unit 22 determines that the level of the reference signal is equal to or higher than the threshold. . Then, when the determination unit 22 determines that the level of the reference signal is less than the threshold, the generation unit 28 may stop the replacement signal generation process.

When the determination unit 22 determines that the level of the reference signal is less than the threshold value, the generation unit 28 stops the replacement signal generation process, thereby reducing the amount of processing calculation of the speech processing device 10 .

When the determination unit 22 determines that the level of the reference signal is equal to or higher than the threshold value, the switching unit 26 replaces the removal signal received from the voice processing unit 24 with the replacement signal received from the generation unit 28 as an output signal to the voice recognition unit. 40. Therefore, when the determination unit 22 determines that the level of the reference signal is equal to or higher than the threshold value, the replacement signal is output to the speech recognition unit 40 instead of the removal signal.

Note that the switching unit 26 may output the replacement signal as an output signal to the speech recognition unit 40 instead of the removal signal during the period in which the determination unit 22 determines that the level of the reference signal is equal to or higher than the threshold. Then, the switching unit 26 may output the removal signal received from the speech processing unit 24 to the speech recognition unit 40 as an output signal during the period when the determination unit 22 determines that the level of the reference signal is less than the threshold value.

In this case, while the level of the reference signal is equal to or higher than the threshold, the replacement signal is output to the speech recognition section 40 as an output signal. Also, during a period in which the level of the reference signal is less than the threshold, the removal signal is output to the speech recognition section 40 as an output signal.

Further, when it is determined that the level of the reference signal is equal to or higher than the threshold, the switching unit 26 outputs the replacement signal instead of the removal signal as an output signal to the speech recognition unit 40 continuously for a predetermined first period of time. good too.

The first time may be determined in advance. For example, at the first time, the output signal output to the speech recognition unit 40 is repeatedly switched between the removal signal and the replacement signal in a short time, and the performance of the speech recognition unit 40 is degraded. A time longer than the continuous output time of the replacement signal to 40 may be determined. Also, for example, for the first time, a value that is equal to or longer than the average utterance period required for uttering one voice command and less than the average utterance period when two voice commands are continuously uttered is set. may Also, the first time may be changed as appropriate according to an operation instruction or the like by the user.

In this case, the replacement signal is output as the output signal to the speech recognition unit 40 continuously for at least the first time from the timing when the level of the reference signal becomes equal to or higher than the threshold. Then, after the first time has elapsed, the removal signal is output to the speech recognition section 40 as an output signal.

Further, when the determination unit 22 determines that the level of the reference signal is equal to or higher than the threshold continuously for a predetermined second time or longer, the switching unit 26 outputs the replacement signal instead of the removal signal as the output signal to the voice recognition unit 40 . You can output to

The second time may be determined in advance. For example, at the second time, the output signal output to the speech recognition unit 40 is repeatedly switched between the removal signal and the replacement signal in a short time, and the performance of the speech recognition unit 40 is degraded. A time longer than the continuous output time of the removal signal or the replacement signal to 40 may be determined. Also, for example, for the second time, a value that is equal to or longer than the average utterance period required for uttering one voice command and less than the average utterance period when two voice commands are continuously uttered is set. may Also, the second time may be changed as appropriate according to an operation instruction or the like by the user.

In this case, when the state in which the level of the reference signal is equal to or higher than the threshold continues for the second time, the replacement signal is output to the speech recognition section 40 as the output signal. Then, when the level of the reference signal is less than the threshold or the duration of the state where the level is equal to or greater than the threshold is less than the second time, the removal signal is output to the speech recognition section 40 as an output signal.

Note that the audio processing unit 24 may always perform removal processing for removing the audio component of the reference signal from the audio signal, but if the determination unit 22 determines that the level of the reference signal is equal to or higher than the threshold, the removal processing is stopped. You may For example, when determining that the level of the reference signal is equal to or higher than the threshold, the determination unit 22 controls the audio processing unit 24 to stop the removal processing.

When it is determined that the level of the reference signal is equal to or higher than the threshold, the audio processing unit 24 stops the removal processing, thereby reducing the amount of processing computation of the audio processing device 10 .

When the level of the reference signal is determined to be equal to or higher than the threshold, the output control unit 29 outputs information indicating that speech recognition is stopped. The output control unit 29 outputs, for example, information indicating that speech recognition is stopped to the display 60 .

As described above, when the level of the reference signal is equal to or higher than the threshold, the replacement signal is output to the speech recognition section 40 as an output signal. Since the replacement signal is at least one of comfort noise and a mute signal, the speech recognition unit 40 does not perform speech recognition while the replacement signal is being accepted. For this reason, for example, in a situation where sound is being emitted in the space inside the vehicle 2 by the loudspeaker SP in accordance with the level of the reproduced signal equal to or higher than the threshold, if the driver hm1 or the like utters a voice command or the like, Even then, the speech recognition unit 40 does not perform speech recognition. Therefore, when it is determined that the level of the reference signal, which is the reproduced signal, is equal to or higher than the threshold, the output control unit 29 outputs information indicating that the speech recognition is stopped so that the user can hear the speech of the speech recognition unit 40. The situation of recognition can be presented easily.

Note that the output target of information by the output control unit 29 is not limited to the display 60 . For example, the output control unit 29 may transmit information indicating that the voice recognition is stopped to an information processing device such as a mobile terminal managed by the pre-registered driver hm1. Further, the output control unit 29 may output information indicating that speech recognition is stopped from the speaker SP. In this case, the level of the reproduced signal of the information indicating that the speech recognition is stopped may be set to a level less than the above threshold.

Next, an example of the flow of information processing executed by the speech processing device 10 of this embodiment will be described.

FIG. 4 is a flowchart showing an example of the flow of information processing executed by the speech processing device 10 of this embodiment.

The voice acquisition unit 20 acquires a voice signal from the microphone MC (step S100).

The determination unit 22 determines whether or not the level of the reference signal, which is the reproduction signal reproduced from the speaker SP, is equal to or higher than the threshold (step S102). If it is determined that the reference signal level is greater than or equal to the threshold (step S102: Yes), the process proceeds to step S104.

In step S104, the determination unit 22 controls the audio processing unit 24 to stop the removal process. By the process of step S104, the audio processing unit 24 stops the removal process.

The generation unit 28 generates a replacement signal that is at least one of the comfort noise and the mute signal, and outputs it to the switching unit 26 (step S106).

The switching unit 26 outputs the replacement signal generated by the generating unit 28 to the speech recognition unit 40 as an output signal (step S108). Since the replacement signal is comfort noise and/or a mute signal, the replacement signal does not include voice commands. Therefore, the voice recognition unit 40 does not recognize voice commands while the replacement signal is being accepted.

The output control unit 29 outputs information indicating that speech recognition is stopped to the display 60 (step S110).

Next, the speech processing device 10 determines whether or not to end the processing (step S112). For example, the sound processing device 10 makes the determination in step S112 by determining whether or not an instruction to cut off power supply to the sound processing device 10 has been given by an operation instruction or the like by the user. If an affirmative determination is made in step S112 (step S112: Yes), the speech processing device 10 ends this routine. When the voice processing device 10 makes a negative determination in step S112 (step S112: No), the process returns to step S100.

On the other hand, if it is determined in step S102 that the level of the reference signal, which is the reproduction signal reproduced from the speaker SP, is less than the threshold (step S102: No), the process proceeds to step S114.

In step S<b>114 , the audio processing unit 24 performs removal processing to generate a removed signal by removing the audio component of the reference signal from the audio signal received from the audio acquisition unit 20 . Note that when the removal process by the audio processing unit 24 has been stopped by the process of step S104, after the determination unit 22 controls the audio processing unit 24 to cancel the stop of the removal process, the audio processing unit 24 should execute the removal process of step S114.

The switching unit 26 outputs the removal signal generated by the speech processing unit 24 to the speech recognition unit 40 as an output signal (step S116). Since the removed signal is a signal obtained by removing the reproduced signal, which is the reference signal, from the audio signal, the removed signal may include a voice command. For this reason, the voice recognition unit 40 is in a state in which voice commands can be recognized during a period in which the removal signal is accepted as an output signal. Then, the process proceeds to step S112.

As described above, the speech processing device 10 of this embodiment includes the speech acquisition unit 20, the determination unit 22, the speech processing unit 24, and the switching unit 26. The audio acquisition unit 20 acquires an audio signal from a microphone MC that picks up spatial audio. The determination unit 22 determines whether or not the level of the reference signal, which is the reproduction signal reproduced from the speaker SP that emits sound in space, is equal to or higher than a threshold. The speech processing unit 24 outputs a removal signal obtained by removing the speech component of the reference signal from the speech signal to the speech recognition unit 40 as an output signal. When the level of the reference signal is determined to be equal to or higher than the threshold, the switching unit 26 outputs a replacement signal, which is at least one of comfort noise and a mute signal, to the speech recognition unit 40 as an output signal instead of the removal signal.

Here, in the conventional technology, a first voice recognition unit recognizes voice picked up by a microphone, a second voice recognition unit recognizes voice output from a speaker, and a second voice recognition unit recognizes voice. A configuration is disclosed that stops recognition by a first speech recognition unit when a speech recognition command is included in the speech recognized by . However, in the prior art, if the voice picked up by the microphone contains noise components such as residual echo components that cannot be removed by an echo canceller or the like, erroneous detection of voice recognition may occur. That is, in the prior art, it was sometimes difficult to suppress erroneous detection of voice recognition. Further, in the prior art, there have been cases where erroneous detection occurs in speech recognition by the first speech recognition unit, depending on the performance of the second speech recognition unit.

On the other hand, in the audio processing apparatus 10 of the present embodiment, when it is determined that the level of the reference signal, which is the reproduced signal, is equal to or higher than the threshold, the audio signal obtained from the microphone MC is replaced with a removed signal obtained by removing the audio component of the reference signal. , the replacement signal, which is at least one of the comfort noise and the mute signal, is output to the speech recognition unit 40 as an output signal. Since the replacement signal is comfort noise and/or a mute signal, the replacement signal does not include voice commands. Therefore, the voice recognition unit 40 does not recognize voice commands while the replacement signal is being accepted.

Therefore, in the audio processing apparatus 10 of the present embodiment, for example, the level of the reproduced signal reproduced from the speaker SP is high, and the audio signal picked up by the microphone MC contains components that cannot be completely canceled by the removal processing. However, it is possible to suppress erroneous detection of voice recognition due to the reproduced signal.

Therefore, the speech processing device 10 of the present embodiment can suppress erroneous detection of speech recognition.

Further, in the audio processing device 10 of the present embodiment, the determination unit 22 determines whether or not the level of the reproduced signal reproduced from the speaker SP, not the level of the audio signal acquired from the microphone MC, is equal to or higher than the threshold. . Therefore, in the speech processing apparatus 10 of the present embodiment, regardless of the level of the speech uttered by the user, if the level of the reproduced signal is less than the threshold, the speech component of the user picked up by the microphone MC can be output to the speech recognition unit 40 as a speech recognition target. Therefore, in addition to the effects described above, the speech processing apparatus 10 of the present embodiment can efficiently perform speech recognition of speech signals including voice commands uttered by the user.

In addition, in the speech processing system 1 of the present embodiment, since speech recognition by the speech recognition unit 40 is not performed on the reproduced signal of the speaker SP, in addition to the above effect, the amount of processing computation of the speech processing system 1 is reduced. reduction can be achieved. Further, in the present embodiment, since speech recognition is not performed on the reproduced signal, erroneous detection of speech recognition can be suppressed regardless of the speech recognition accuracy of the speech recognition unit 40 .

In addition, in this embodiment, the voice processing system 1 has been described as being mounted in the vehicle 2 as an example. However, the voice processing system 1 is not limited to being mounted on the vehicle 2 as long as it is arranged in an arbitrary space for voice processing.

Although the embodiment has been described above, the embodiment is presented as an example and is not intended to limit the scope of the invention. The novel embodiments described above can be embodied in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. The above embodiments are included in the scope or gist of the invention, and are included in the scope of the invention described in the claims and equivalents thereof.

1 Speech processing system 10 Speech processing device 20 Speech acquisition unit 22 Judgment unit 24 Speech processing unit 26 Switching unit 28 Generation unit 40 Speech recognition unit 50 Electronic device 60 Display MC Microphone SP Speaker

Claims (10)

an audio acquisition unit that acquires an audio signal from a microphone that picks up spatial audio;
a determination unit that determines whether a level of a reference signal, which is a reproduction signal reproduced from a speaker that emits sound in the space, is equal to or higher than a threshold;
a speech processing unit that outputs a removal signal obtained by removing the speech component of the reference signal from the speech signal to a speech recognition unit as an output signal;
a switching unit that outputs a replacement signal, which is at least one of comfort noise and a mute signal, as the output signal to the speech recognition unit instead of the removal signal when the level of the reference signal is determined to be equal to or higher than the threshold;
A speech processing device comprising: The determination unit
When the level of the reference signal is determined to be equal to or higher than the threshold, controlling the audio processing unit to stop removal processing for removing the audio component of the reference signal from the audio signal;
The audio processing device according to claim 1. Further comprising a generation unit that generates the replacement signal,
The generating unit
generating the replacement signal that is the comfort noise according to the noise level included in the audio signal immediately before it is determined to be equal to or greater than the threshold;
3. The audio processing device according to claim 1 or 2. The switching unit is
outputting the replacement signal as the output signal to the speech recognition unit in place of the removal signal during a period in which the level of the reference signal is determined to be equal to or higher than the threshold;
The audio processing device according to any one of claims 1 to 3. The switching unit is
when the level of the reference signal is determined to be equal to or higher than the threshold, continuously outputting the replacement signal as the output signal instead of the removal signal to the speech recognition unit for a predetermined first time;
The audio processing device according to any one of claims 1 to 3. The switching unit is
3. Outputting the replacement signal to the speech recognition unit as the output signal in place of the removal signal when the level of the reference signal is continuously determined to be the threshold value or more for a predetermined second time period or longer. The audio processing device according to any one of claims 1 to 3. an output control unit that outputs information indicating that speech recognition is stopped when the level of the reference signal is determined to be equal to or higher than the threshold;
The audio processing device according to any one of claims 1 to 6, further comprising: A speech processing method executed by a speech processing device,
acquiring an audio signal from a microphone that picks up spatial audio;
a step of determining whether the level of a reference signal, which is a reproduction signal reproduced from a speaker emitting sound in the space, is equal to or higher than a threshold;
a step of outputting, as an output signal, a removed signal obtained by removing the audio component of the reference signal from the audio signal to a speech recognition unit;
a step of outputting a replacement signal, which is at least one of comfort noise and a mute signal, as the output signal to the speech recognition unit in place of the removal signal when the level of the reference signal is determined to be equal to or higher than the threshold;
audio processing methods, including acquiring an audio signal from a microphone that picks up spatial audio;
a step of determining whether the level of a reference signal, which is a reproduction signal reproduced from a speaker emitting sound in the space, is equal to or higher than a threshold;
a step of outputting, as an output signal, a removed signal obtained by removing the audio component of the reference signal from the audio signal to a speech recognition unit;
a step of outputting a replacement signal, which is at least one of comfort noise and a mute signal, as the output signal to the speech recognition unit in place of the removal signal when the level of the reference signal is determined to be equal to or higher than the threshold;
A sound processing program that causes a computer to execute A voice processing system comprising a voice processing device, a microphone that collects voice in a space, a speaker that outputs sound to the space, and a voice recognition unit that recognizes voice,
The audio processing device is
an audio acquisition unit that acquires an audio signal from the microphone;
a determination unit that determines whether a level of a reference signal, which is a reproduction signal reproduced from the speaker, is equal to or higher than a threshold;
a speech processing unit that outputs a removal signal obtained by removing the speech component of the reference signal from the speech signal to the speech recognition unit as an output signal;
a switching unit that outputs a replacement signal, which is at least one of comfort noise and a mute signal, as the output signal to the speech recognition unit instead of the removal signal when the level of the reference signal is determined to be equal to or higher than the threshold;
An audio processing system comprising:

Images